Model

The previous section described my overall theory about analogical problem solving and the role of visual analogy. In this section I will flesh out the parts of the theory that I will evaluate.

According to my theory, episodes of problem solving can be represented as a sequence of knowledge states in the problem solving procedure, connected by manipulations between them. Manipulations are operations the agent can take on the system to change it. These are distinguished from simulation events, which occur in the system on their own. For example, the workings of a clock are simulation events, and winding the clock is a manipulation. Applying a particular manipulation to a knowledge state results in another, changed knowledge state. They are states in the problem solving process.

Problems and solution procedures can be represented non-visually and visually. In the non-visual representation, the knowledge states are called nv-states, and the manipulations are called actions. Both actions and nv-states can be transformed into visual representations. This is done using the Cognitive Visual Language, Covlan, which provides the vocabulary and processes for turning non-visual physical system representations into visual ones. Covlan knowledge states and manipulations are called s-images and transformations, respectively.

Unsolved target problems are represented as single nv-states or s-images.

Table 1
	Knowledge State	Entity	Manipulation
non-visual	nv-state	object	action
Covlan	s-image	element	transformation

First I will describe the representation language, then the inference and control of the theory.